ai-regulation news, stories and analysis

Treemap of five AI danger domains Bill Gates says we have crossed, sized by relative emphasis in his essay. Bio-capabilities at 50, cyber-capabilities at 35, job-market destruction at 30, psychosocial at 20, and control at 15. AI danger thresholds.

Bill Gates says we crossed AI danger thresholds. Now what?

Bill Gates says we have crossed AI danger thresholds in bio, cyber, jobs, and control. He now calls bioterrorism 50 times more likely than a natural pandemic.

Lars Cornelissen · Aug 26, 2026

Radar chart comparing Geoffrey Hinton, Andrew Ng, and Fei-Fei Li on five dimensions of the open-weight debate: openness advocacy, risk concern, regulation support, China concern, and acceptance of permanence. Hinton scores 9 on risk concern and 10 on acceptance. Ng scores 9 on openness and 9 on China concern. Li scores 7 on openness and 5 on risk concern.

AI

Open-weight AI models: Hinton says the battle is lost

Open-weight AI models are now permanent, three AI pioneers agreed at Ai4. Hinton conceded the battle is lost while Ng and Li called for openness with nuance.

Lars Cornelissen · Aug 13, 2026

Gantt chart of EU AI Act enforcement phases. Prohibited practices began February 2025 and remain ongoing. Transparency rules took effect August 2026 with a marking grace period ending December 2026. High-risk obligations begin December 2027 through August 2028.

AI

EU AI Act transparency rules take effect, December deadline

EU AI Act transparency rules took effect August 2, 2026, mandating chatbot disclosure and machine-readable AI content marking. Fines reach €15M or 3% of global turnover.

Lars Cornelissen · Aug 4, 2026